arch/ia64/kernel/process.c

   1 /*
   2  * Architecture-specific setup.
   3  *
   4  * Copyright (C) 1998-2003 Hewlett-Packard Co
   5  *      David Mosberger-Tang <davidm@hpl.hp.com>
   6  * 04/11/17 Ashok Raj   <ashok.raj@intel.com> Added CPU Hotplug Support
   7  *
   8  * 2005-10-07 Keith Owens <kaos@sgi.com>
   9  *            Add notify_die() hooks.
  10  */
  11 #include <linux/cpu.h>
  12 #include <linux/pm.h>
  13 #include <linux/elf.h>
  14 #include <linux/errno.h>
  15 #include <linux/kallsyms.h>
  16 #include <linux/kernel.h>
  17 #include <linux/mm.h>
  18 #include <linux/module.h>
  19 #include <linux/notifier.h>
  20 #include <linux/personality.h>
  21 #include <linux/sched.h>
  22 #include <linux/slab.h>
  23 #include <linux/stddef.h>
  24 #include <linux/thread_info.h>
  25 #include <linux/unistd.h>
  26 #include <linux/efi.h>
  27 #include <linux/interrupt.h>
  28 #include <linux/delay.h>
  29 #include <linux/kdebug.h>
  30 #include <linux/utsname.h>
  31
  32 #include <asm/cpu.h>
  33 #include <asm/delay.h>
  34 #include <asm/elf.h>
  35 #include <asm/ia32.h>
  36 #include <asm/irq.h>
  37 #include <asm/kexec.h>
  38 #include <asm/pgalloc.h>
  39 #include <asm/processor.h>
  40 #include <asm/sal.h>
  41 #include <asm/tlbflush.h>
  42 #include <asm/uaccess.h>
  43 #include <asm/unwind.h>
  44 #include <asm/user.h>
  45
  46 #include "entry.h"
  47
  48 #ifdef CONFIG_PERFMON
  49 # include <asm/perfmon.h>
  50 #endif
  51
  52 #include "sigframe.h"
  53
  54 void (*ia64_mark_idle)(int);
  55 static DEFINE_PER_CPU(unsigned int, cpu_idle_state);
  56
  57 unsigned long boot_option_idle_override = 0;
  58 EXPORT_SYMBOL(boot_option_idle_override);
  59
  60 void
  61 ia64_do_show_stack (struct unw_frame_info *info, void *arg)
  62 {
  63         unsigned long ip, sp, bsp;
  64         char buf[128];                  /* don't make it so big that it overflows the stack! */
  65
  66         printk("\nCall Trace:\n");
  67         do {
  68                 unw_get_ip(info, &ip);
  69                 if (ip == 0)
  70                         break;
  71
  72                 unw_get_sp(info, &sp);
  73                 unw_get_bsp(info, &bsp);
  74                 snprintf(buf, sizeof(buf),
  75                          " [<%016lx>] %%s\n"
  76                          "                                sp=%016lx bsp=%016lx\n",
  77                          ip, sp, bsp);
  78                 print_symbol(buf, ip);
  79         } while (unw_unwind(info) >= 0);
  80 }
  81
  82 void
  83 show_stack (struct task_struct *task, unsigned long *sp)
  84 {
  85         if (!task)
  86                 unw_init_running(ia64_do_show_stack, NULL);
  87         else {
  88                 struct unw_frame_info info;
  89
  90                 unw_init_from_blocked_task(&info, task);
  91                 ia64_do_show_stack(&info, NULL);
  92         }
  93 }
  94
  95 void
  96 dump_stack (void)
  97 {
  98         show_stack(NULL, NULL);
  99 }
 100
 101 EXPORT_SYMBOL(dump_stack);
 102
 103 void
 104 show_regs (struct pt_regs *regs)
 105 {
 106         unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
 107
 108         print_modules();
 109         printk("\nPid: %d, CPU %d, comm: %20s\n", task_pid_nr(current),
 110                         smp_processor_id(), current->comm);
 111         printk("psr : %016lx ifs : %016lx ip  : [<%016lx>]    %s (%s)\n",
 112                regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(),
 113                init_utsname()->release);
 114         print_symbol("ip is at %s\n", ip);
 115         printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
 116                regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
 117         printk("rnat: %016lx bsps: %016lx pr  : %016lx\n",
 118                regs->ar_rnat, regs->ar_bspstore, regs->pr);
 119         printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
 120                regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
 121         printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
 122         printk("b0  : %016lx b6  : %016lx b7  : %016lx\n", regs->b0, regs->b6, regs->b7);
 123         printk("f6  : %05lx%016lx f7  : %05lx%016lx\n",
 124                regs->f6.u.bits[1], regs->f6.u.bits[0],
 125                regs->f7.u.bits[1], regs->f7.u.bits[0]);
 126         printk("f8  : %05lx%016lx f9  : %05lx%016lx\n",
 127                regs->f8.u.bits[1], regs->f8.u.bits[0],
 128                regs->f9.u.bits[1], regs->f9.u.bits[0]);
 129         printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
 130                regs->f10.u.bits[1], regs->f10.u.bits[0],
 131                regs->f11.u.bits[1], regs->f11.u.bits[0]);
 132
 133         printk("r1  : %016lx r2  : %016lx r3  : %016lx\n", regs->r1, regs->r2, regs->r3);
 134         printk("r8  : %016lx r9  : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
 135         printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
 136         printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
 137         printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
 138         printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
 139         printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
 140         printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
 141         printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
 142
 143         if (user_mode(regs)) {
 144                 /* print the stacked registers */
 145                 unsigned long val, *bsp, ndirty;
 146                 int i, sof, is_nat = 0;
 147
 148                 sof = regs->cr_ifs & 0x7f;      /* size of frame */
 149                 ndirty = (regs->loadrs >> 19);
 150                 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
 151                 for (i = 0; i < sof; ++i) {
 152                         get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
 153                         printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
 154                                ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
 155                 }
 156         } else
 157                 show_stack(NULL, NULL);
 158 }
 159
 160 void
 161 do_notify_resume_user (sigset_t *unused, struct sigscratch *scr, long in_syscall)
 162 {
 163         if (fsys_mode(current, &scr->pt)) {
 164                 /* defer signal-handling etc. until we return to privilege-level 0.  */
 165                 if (!ia64_psr(&scr->pt)->lp)
 166                         ia64_psr(&scr->pt)->lp = 1;
 167                 return;
 168         }
 169
 170 #ifdef CONFIG_PERFMON
 171         if (current->thread.pfm_needs_checking)
 172                 pfm_handle_work();
 173 #endif
 174
 175         /* deal with pending signal delivery */
 176         if (test_thread_flag(TIF_SIGPENDING)||test_thread_flag(TIF_RESTORE_SIGMASK))
 177                 ia64_do_signal(scr, in_syscall);
 178 }
 179
 180 static int pal_halt        = 1;
 181 static int can_do_pal_halt = 1;
 182
 183 static int __init nohalt_setup(char * str)
 184 {
 185         pal_halt = can_do_pal_halt = 0;
 186         return 1;
 187 }
 188 __setup("nohalt", nohalt_setup);
 189
 190 void
 191 update_pal_halt_status(int status)
 192 {
 193         can_do_pal_halt = pal_halt && status;
 194 }
 195
 196 /*
 197  * We use this if we don't have any better idle routine..
 198  */
 199 void
 200 default_idle (void)
 201 {
 202         local_irq_enable();
 203         while (!need_resched()) {
 204                 if (can_do_pal_halt) {
 205                         local_irq_disable();
 206                         if (!need_resched()) {
 207                                 safe_halt();
 208                         }
 209                         local_irq_enable();
 210                 } else
 211                         cpu_relax();
 212         }
 213 }
 214
 215 #ifdef CONFIG_HOTPLUG_CPU
 216 /* We don't actually take CPU down, just spin without interrupts. */
 217 static inline void play_dead(void)
 218 {
 219         extern void ia64_cpu_local_tick (void);
 220         unsigned int this_cpu = smp_processor_id();
 221
 222         /* Ack it */
 223         __get_cpu_var(cpu_state) = CPU_DEAD;
 224
 225         max_xtp();
 226         local_irq_disable();
 227         idle_task_exit();
 228         ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
 229         /*
 230          * The above is a point of no-return, the processor is
 231          * expected to be in SAL loop now.
 232          */
 233         BUG();
 234 }
 235 #else
 236 static inline void play_dead(void)
 237 {
 238         BUG();
 239 }
 240 #endif /* CONFIG_HOTPLUG_CPU */
 241
 242 void cpu_idle_wait(void)
 243 {
 244         unsigned int cpu, this_cpu = get_cpu();
 245         cpumask_t map;
 246         cpumask_t tmp = current->cpus_allowed;
 247
 248         set_cpus_allowed(current, cpumask_of_cpu(this_cpu));
 249         put_cpu();
 250
 251         cpus_clear(map);
 252         for_each_online_cpu(cpu) {
 253                 per_cpu(cpu_idle_state, cpu) = 1;
 254                 cpu_set(cpu, map);
 255         }
 256
 257         __get_cpu_var(cpu_idle_state) = 0;
 258
 259         wmb();
 260         do {
 261                 ssleep(1);
 262                 for_each_online_cpu(cpu) {
 263                         if (cpu_isset(cpu, map) && !per_cpu(cpu_idle_state, cpu))
 264                                 cpu_clear(cpu, map);
 265                 }
 266                 cpus_and(map, map, cpu_online_map);
 267         } while (!cpus_empty(map));
 268         set_cpus_allowed(current, tmp);
 269 }
 270 EXPORT_SYMBOL_GPL(cpu_idle_wait);
 271
 272 void __attribute__((noreturn))
 273 cpu_idle (void)
 274 {
 275         void (*mark_idle)(int) = ia64_mark_idle;
 276         int cpu = smp_processor_id();
 277
 278         /* endless idle loop with no priority at all */
 279         while (1) {
 280                 if (can_do_pal_halt) {
 281                         current_thread_info()->status &= ~TS_POLLING;
 282                         /*
 283                          * TS_POLLING-cleared state must be visible before we
 284                          * test NEED_RESCHED:
 285                          */
 286                         smp_mb();
 287                 } else {
 288                         current_thread_info()->status |= TS_POLLING;
 289                 }
 290
 291                 if (!need_resched()) {
 292                         void (*idle)(void);
 293 #ifdef CONFIG_SMP
 294                         min_xtp();
 295 #endif
 296                         if (__get_cpu_var(cpu_idle_state))
 297                                 __get_cpu_var(cpu_idle_state) = 0;
 298
 299                         rmb();
 300                         if (mark_idle)
 301                                 (*mark_idle)(1);
 302
 303                         idle = pm_idle;
 304                         if (!idle)
 305                                 idle = default_idle;
 306                         (*idle)();
 307                         if (mark_idle)
 308                                 (*mark_idle)(0);
 309 #ifdef CONFIG_SMP
 310                         normal_xtp();
 311 #endif
 312                 }
 313                 preempt_enable_no_resched();
 314                 schedule();
 315                 preempt_disable();
 316                 check_pgt_cache();
 317                 if (cpu_is_offline(cpu))
 318                         play_dead();
 319         }
 320 }
 321
 322 void
 323 ia64_save_extra (struct task_struct *task)
 324 {
 325 #ifdef CONFIG_PERFMON
 326         unsigned long info;
 327 #endif
 328
 329         if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
 330                 ia64_save_debug_regs(&task->thread.dbr[0]);
 331
 332 #ifdef CONFIG_PERFMON
 333         if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
 334                 pfm_save_regs(task);
 335
 336         info = __get_cpu_var(pfm_syst_info);
 337         if (info & PFM_CPUINFO_SYST_WIDE)
 338                 pfm_syst_wide_update_task(task, info, 0);
 339 #endif
 340
 341 #ifdef CONFIG_IA32_SUPPORT
 342         if (IS_IA32_PROCESS(task_pt_regs(task)))
 343                 ia32_save_state(task);
 344 #endif
 345 }
 346
 347 void
 348 ia64_load_extra (struct task_struct *task)
 349 {
 350 #ifdef CONFIG_PERFMON
 351         unsigned long info;
 352 #endif
 353
 354         if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
 355                 ia64_load_debug_regs(&task->thread.dbr[0]);
 356
 357 #ifdef CONFIG_PERFMON
 358         if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
 359                 pfm_load_regs(task);
 360
 361         info = __get_cpu_var(pfm_syst_info);
 362         if (info & PFM_CPUINFO_SYST_WIDE)
 363                 pfm_syst_wide_update_task(task, info, 1);
 364 #endif
 365
 366 #ifdef CONFIG_IA32_SUPPORT
 367         if (IS_IA32_PROCESS(task_pt_regs(task)))
 368                 ia32_load_state(task);
 369 #endif
 370 }
 371
 372 /*
 373  * Copy the state of an ia-64 thread.
 374  *
 375  * We get here through the following  call chain:
 376  *
 377  *      from user-level:        from kernel:
 378  *
 379  *      <clone syscall>         <some kernel call frames>
 380  *      sys_clone                  :
 381  *      do_fork                 do_fork
 382  *      copy_thread             copy_thread
 383  *
 384  * This means that the stack layout is as follows:
 385  *
 386  *      +---------------------+ (highest addr)
 387  *      |   struct pt_regs    |
 388  *      +---------------------+
 389  *      | struct switch_stack |
 390  *      +---------------------+
 391  *      |                     |
 392  *      |    memory stack     |
 393  *      |                     | <-- sp (lowest addr)
 394  *      +---------------------+
 395  *
 396  * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
 397  * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
 398  * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
 399  * pt_regs structure in the parent is congruent to that of the child, modulo 512.  Since
 400  * the stack is page aligned and the page size is at least 4KB, this is always the case,
 401  * so there is nothing to worry about.
 402  */
 403 int
 404 copy_thread (int nr, unsigned long clone_flags,
 405              unsigned long user_stack_base, unsigned long user_stack_size,
 406              struct task_struct *p, struct pt_regs *regs)
 407 {
 408         extern char ia64_ret_from_clone, ia32_ret_from_clone;
 409         struct switch_stack *child_stack, *stack;
 410         unsigned long rbs, child_rbs, rbs_size;
 411         struct pt_regs *child_ptregs;
 412         int retval = 0;
 413
 414 #ifdef CONFIG_SMP
 415         /*
 416          * For SMP idle threads, fork_by_hand() calls do_fork with
 417          * NULL regs.
 418          */
 419         if (!regs)
 420                 return 0;
 421 #endif
 422
 423         stack = ((struct switch_stack *) regs) - 1;
 424
 425         child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
 426         child_stack = (struct switch_stack *) child_ptregs - 1;
 427
 428         /* copy parent's switch_stack & pt_regs to child: */
 429         memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
 430
 431         rbs = (unsigned long) current + IA64_RBS_OFFSET;
 432         child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
 433         rbs_size = stack->ar_bspstore - rbs;
 434
 435         /* copy the parent's register backing store to the child: */
 436         memcpy((void *) child_rbs, (void *) rbs, rbs_size);
 437
 438         if (likely(user_mode(child_ptregs))) {
 439                 if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs))
 440                         child_ptregs->r13 = regs->r16;  /* see sys_clone2() in entry.S */
 441                 if (user_stack_base) {
 442                         child_ptregs->r12 = user_stack_base + user_stack_size - 16;
 443                         child_ptregs->ar_bspstore = user_stack_base;
 444                         child_ptregs->ar_rnat = 0;
 445                         child_ptregs->loadrs = 0;
 446                 }
 447         } else {
 448                 /*
 449                  * Note: we simply preserve the relative position of
 450                  * the stack pointer here.  There is no need to
 451                  * allocate a scratch area here, since that will have
 452                  * been taken care of by the caller of sys_clone()
 453                  * already.
 454                  */
 455                 child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
 456                 child_ptregs->r13 = (unsigned long) p;          /* set `current' pointer */
 457         }
 458         child_stack->ar_bspstore = child_rbs + rbs_size;
 459         if (IS_IA32_PROCESS(regs))
 460                 child_stack->b0 = (unsigned long) &ia32_ret_from_clone;
 461         else
 462                 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
 463
 464         /* copy parts of thread_struct: */
 465         p->thread.ksp = (unsigned long) child_stack - 16;
 466
 467         /* stop some PSR bits from being inherited.
 468          * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
 469          * therefore we must specify them explicitly here and not include them in
 470          * IA64_PSR_BITS_TO_CLEAR.
 471          */
 472         child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
 473                                  & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
 474
 475         /*
 476          * NOTE: The calling convention considers all floating point
 477          * registers in the high partition (fph) to be scratch.  Since
 478          * the only way to get to this point is through a system call,
 479          * we know that the values in fph are all dead.  Hence, there
 480          * is no need to inherit the fph state from the parent to the
 481          * child and all we have to do is to make sure that
 482          * IA64_THREAD_FPH_VALID is cleared in the child.
 483          *
 484          * XXX We could push this optimization a bit further by
 485          * clearing IA64_THREAD_FPH_VALID on ANY system call.
 486          * However, it's not clear this is worth doing.  Also, it
 487          * would be a slight deviation from the normal Linux system
 488          * call behavior where scratch registers are preserved across
 489          * system calls (unless used by the system call itself).
 490          */
 491 #       define THREAD_FLAGS_TO_CLEAR    (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
 492                                          | IA64_THREAD_PM_VALID)
 493 #       define THREAD_FLAGS_TO_SET      0
 494         p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
 495                            | THREAD_FLAGS_TO_SET);
 496         ia64_drop_fpu(p);       /* don't pick up stale state from a CPU's fph */
 497 #ifdef CONFIG_IA32_SUPPORT
 498         /*
 499          * If we're cloning an IA32 task then save the IA32 extra
 500          * state from the current task to the new task
 501          */
 502         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 503                 ia32_save_state(p);
 504                 if (clone_flags & CLONE_SETTLS)
 505                         retval = ia32_clone_tls(p, child_ptregs);
 506
 507                 /* Copy partially mapped page list */
 508                 if (!retval)
 509                         retval = ia32_copy_ia64_partial_page_list(p,
 510                                                                 clone_flags);
 511         }
 512 #endif
 513
 514 #ifdef CONFIG_PERFMON
 515         if (current->thread.pfm_context)
 516                 pfm_inherit(p, child_ptregs);
 517 #endif
 518         return retval;
 519 }
 520
 521 static void
 522 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
 523 {
 524         unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
 525         unsigned long uninitialized_var(ip);    /* GCC be quiet */
 526         elf_greg_t *dst = arg;
 527         struct pt_regs *pt;
 528         char nat;
 529         int i;
 530
 531         memset(dst, 0, sizeof(elf_gregset_t));  /* don't leak any kernel bits to user-level */
 532
 533         if (unw_unwind_to_user(info) < 0)
 534                 return;
 535
 536         unw_get_sp(info, &sp);
 537         pt = (struct pt_regs *) (sp + 16);
 538
 539         urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
 540
 541         if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
 542                 return;
 543
 544         ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
 545                   &ar_rnat);
 546
 547         /*
 548          * coredump format:
 549          *      r0-r31
 550          *      NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
 551          *      predicate registers (p0-p63)
 552          *      b0-b7
 553          *      ip cfm user-mask
 554          *      ar.rsc ar.bsp ar.bspstore ar.rnat
 555          *      ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
 556          */
 557
 558         /* r0 is zero */
 559         for (i = 1, mask = (1UL << i); i < 32; ++i) {
 560                 unw_get_gr(info, i, &dst[i], &nat);
 561                 if (nat)
 562                         nat_bits |= mask;
 563                 mask <<= 1;
 564         }
 565         dst[32] = nat_bits;
 566         unw_get_pr(info, &dst[33]);
 567
 568         for (i = 0; i < 8; ++i)
 569                 unw_get_br(info, i, &dst[34 + i]);
 570
 571         unw_get_rp(info, &ip);
 572         dst[42] = ip + ia64_psr(pt)->ri;
 573         dst[43] = cfm;
 574         dst[44] = pt->cr_ipsr & IA64_PSR_UM;
 575
 576         unw_get_ar(info, UNW_AR_RSC, &dst[45]);
 577         /*
 578          * For bsp and bspstore, unw_get_ar() would return the kernel
 579          * addresses, but we need the user-level addresses instead:
 580          */
 581         dst[46] = urbs_end;     /* note: by convention PT_AR_BSP points to the end of the urbs! */
 582         dst[47] = pt->ar_bspstore;
 583         dst[48] = ar_rnat;
 584         unw_get_ar(info, UNW_AR_CCV, &dst[49]);
 585         unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
 586         unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
 587         dst[52] = pt->ar_pfs;   /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
 588         unw_get_ar(info, UNW_AR_LC, &dst[53]);
 589         unw_get_ar(info, UNW_AR_EC, &dst[54]);
 590         unw_get_ar(info, UNW_AR_CSD, &dst[55]);
 591         unw_get_ar(info, UNW_AR_SSD, &dst[56]);
 592 }
 593
 594 void
 595 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
 596 {
 597         elf_fpreg_t *dst = arg;
 598         int i;
 599
 600         memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */
 601
 602         if (unw_unwind_to_user(info) < 0)
 603                 return;
 604
 605         /* f0 is 0.0, f1 is 1.0 */
 606
 607         for (i = 2; i < 32; ++i)
 608                 unw_get_fr(info, i, dst + i);
 609
 610         ia64_flush_fph(task);
 611         if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
 612                 memcpy(dst + 32, task->thread.fph, 96*16);
 613 }
 614
 615 void
 616 do_copy_regs (struct unw_frame_info *info, void *arg)
 617 {
 618         do_copy_task_regs(current, info, arg);
 619 }
 620
 621 void
 622 do_dump_fpu (struct unw_frame_info *info, void *arg)
 623 {
 624         do_dump_task_fpu(current, info, arg);
 625 }
 626
 627 int
 628 dump_task_regs(struct task_struct *task, elf_gregset_t *regs)
 629 {
 630         struct unw_frame_info tcore_info;
 631
 632         if (current == task) {
 633                 unw_init_running(do_copy_regs, regs);
 634         } else {
 635                 memset(&tcore_info, 0, sizeof(tcore_info));
 636                 unw_init_from_blocked_task(&tcore_info, task);
 637                 do_copy_task_regs(task, &tcore_info, regs);
 638         }
 639         return 1;
 640 }
 641
 642 void
 643 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
 644 {
 645         unw_init_running(do_copy_regs, dst);
 646 }
 647
 648 int
 649 dump_task_fpu (struct task_struct *task, elf_fpregset_t *dst)
 650 {
 651         struct unw_frame_info tcore_info;
 652
 653         if (current == task) {
 654                 unw_init_running(do_dump_fpu, dst);
 655         } else {
 656                 memset(&tcore_info, 0, sizeof(tcore_info));
 657                 unw_init_from_blocked_task(&tcore_info, task);
 658                 do_dump_task_fpu(task, &tcore_info, dst);
 659         }
 660         return 1;
 661 }
 662
 663 int
 664 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
 665 {
 666         unw_init_running(do_dump_fpu, dst);
 667         return 1;       /* f0-f31 are always valid so we always return 1 */
 668 }
 669
 670 long
 671 sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp,
 672             struct pt_regs *regs)
 673 {
 674         char *fname;
 675         int error;
 676
 677         fname = getname(filename);
 678         error = PTR_ERR(fname);
 679         if (IS_ERR(fname))
 680                 goto out;
 681         error = do_execve(fname, argv, envp, regs);
 682         putname(fname);
 683 out:
 684         return error;
 685 }
 686
 687 pid_t
 688 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
 689 {
 690         extern void start_kernel_thread (void);
 691         unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
 692         struct {
 693                 struct switch_stack sw;
 694                 struct pt_regs pt;
 695         } regs;
 696
 697         memset(&regs, 0, sizeof(regs));
 698         regs.pt.cr_iip = helper_fptr[0];        /* set entry point (IP) */
 699         regs.pt.r1 = helper_fptr[1];            /* set GP */
 700         regs.pt.r9 = (unsigned long) fn;        /* 1st argument */
 701         regs.pt.r11 = (unsigned long) arg;      /* 2nd argument */
 702         /* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
 703         regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
 704         regs.pt.cr_ifs = 1UL << 63;             /* mark as valid, empty frame */
 705         regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
 706         regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
 707         regs.sw.pr = (1 << PRED_KERNEL_STACK);
 708         return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
 709 }
 710 EXPORT_SYMBOL(kernel_thread);
 711
 712 /* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
 713 int
 714 kernel_thread_helper (int (*fn)(void *), void *arg)
 715 {
 716 #ifdef CONFIG_IA32_SUPPORT
 717         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 718                 /* A kernel thread is always a 64-bit process. */
 719                 current->thread.map_base  = DEFAULT_MAP_BASE;
 720                 current->thread.task_size = DEFAULT_TASK_SIZE;
 721                 ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
 722                 ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
 723         }
 724 #endif
 725         return (*fn)(arg);
 726 }
 727
 728 /*
 729  * Flush thread state.  This is called when a thread does an execve().
 730  */
 731 void
 732 flush_thread (void)
 733 {
 734         /* drop floating-point and debug-register state if it exists: */
 735         current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
 736         ia64_drop_fpu(current);
 737 #ifdef CONFIG_IA32_SUPPORT
 738         if (IS_IA32_PROCESS(task_pt_regs(current))) {
 739                 ia32_drop_ia64_partial_page_list(current);
 740                 current->thread.task_size = IA32_PAGE_OFFSET;
 741                 set_fs(USER_DS);
 742                 memset(current->thread.tls_array, 0, sizeof(current->thread.tls_array));
 743         }
 744 #endif
 745 }
 746
 747 /*
 748  * Clean up state associated with current thread.  This is called when
 749  * the thread calls exit().
 750  */
 751 void
 752 exit_thread (void)
 753 {
 754
 755         ia64_drop_fpu(current);
 756 #ifdef CONFIG_PERFMON
 757        /* if needed, stop monitoring and flush state to perfmon context */
 758         if (current->thread.pfm_context)
 759                 pfm_exit_thread(current);
 760
 761         /* free debug register resources */
 762         if (current->thread.flags & IA64_THREAD_DBG_VALID)
 763                 pfm_release_debug_registers(current);
 764 #endif
 765         if (IS_IA32_PROCESS(task_pt_regs(current)))
 766                 ia32_drop_ia64_partial_page_list(current);
 767 }
 768
 769 unsigned long
 770 get_wchan (struct task_struct *p)
 771 {
 772         struct unw_frame_info info;
 773         unsigned long ip;
 774         int count = 0;
 775
 776         if (!p || p == current || p->state == TASK_RUNNING)
 777                 return 0;
 778
 779         /*
 780          * Note: p may not be a blocked task (it could be current or
 781          * another process running on some other CPU.  Rather than
 782          * trying to determine if p is really blocked, we just assume
 783          * it's blocked and rely on the unwind routines to fail
 784          * gracefully if the process wasn't really blocked after all.
 785          * --davidm 99/12/15
 786          */
 787         unw_init_from_blocked_task(&info, p);
 788         do {
 789                 if (p->state == TASK_RUNNING)
 790                         return 0;
 791                 if (unw_unwind(&info) < 0)
 792                         return 0;
 793                 unw_get_ip(&info, &ip);
 794                 if (!in_sched_functions(ip))
 795                         return ip;
 796         } while (count++ < 16);
 797         return 0;
 798 }
 799
 800 void
 801 cpu_halt (void)
 802 {
 803         pal_power_mgmt_info_u_t power_info[8];
 804         unsigned long min_power;
 805         int i, min_power_state;
 806
 807         if (ia64_pal_halt_info(power_info) != 0)
 808                 return;
 809
 810         min_power_state = 0;
 811         min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
 812         for (i = 1; i < 8; ++i)
 813                 if (power_info[i].pal_power_mgmt_info_s.im
 814                     && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
 815                         min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
 816                         min_power_state = i;
 817                 }
 818
 819         while (1)
 820                 ia64_pal_halt(min_power_state);
 821 }
 822
 823 void machine_shutdown(void)
 824 {
 825 #ifdef CONFIG_HOTPLUG_CPU
 826         int cpu;
 827
 828         for_each_online_cpu(cpu) {
 829                 if (cpu != smp_processor_id())
 830                         cpu_down(cpu);
 831         }
 832 #endif
 833 #ifdef CONFIG_KEXEC
 834         kexec_disable_iosapic();
 835 #endif
 836 }
 837
 838 void
 839 machine_restart (char *restart_cmd)
 840 {
 841         (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
 842         (*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
 843 }
 844
 845 void
 846 machine_halt (void)
 847 {
 848         (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
 849         cpu_halt();
 850 }
 851
 852 void
 853 machine_power_off (void)
 854 {
 855         if (pm_power_off)
 856                 pm_power_off();
 857         machine_halt();
 858 }
 859